An issue that has received a great deal of attention in recent years is the relative importance of age and time to death in determining health care costs. It is well-known that health care costs rise sharply at the end of life, especially in the year before death. This has led many to the view that population aging, insofar as it is a consequence of increased life expectancy, might not produce proportionate increases in Medicare costs, because the costs associated with the extra years lived may simply be experienced later in life. However, all studies of the age-versus-time-to-death question have used a model in which, conditional upon measured attributes, there is a single population-level pathway of expected Medicare costs. Yet there are strong reasons to suppose that there are several prototypical mean pathways of Medicare costs as individuals approach death. We will build a short-term Medicare-cost forecasting model that recognizes the existence of a set of distinctive, but unobserved, cost-trajectory types. Specifically, this project will: " Estimate an integrated model of mortality and Medicare costs-both in the aggregate and within major categories such as inpatient care, outpatient care, SNF usage and so on-using a generalization of latent-class trajectory models and applying longitudinal data from the National Long Term Care Survey and the Cardiovascular Health Study, both of which have been linked to continuous Medicare claims records;" Produce new estimates of the relative effects of age and time to death on health care costs, while controlling for diagnoses and selected service-use indicators;and " Use microsimulation to compute complete-cohort estimates of total and component Medicare costs, for samples of synthetic observations created from the 2 data files used in the analysis. PUBLIC HEALTH RELEVANCE: Medicare costs provide a measure of our commitment to maintaining and improving public health, as well as an index of our success in achieving public-health goals with respect to the nation's older citizens. Understanding the ways that ongoing changes in the population-greater longevity, a trend towards lower rates of disability at older ages, and the transition into late life of members of the "Baby Boom"-are likely to influence future Medicare costs is vital to our capacity to plan for the future. This project takes a new look at the question of whether age, or nearness to death, is mainly responsible for the observed patterns of Medicare costs that rise with age;the ultimate goal of the project is to produce improved means of forecasting future Medicare costs.